In robotically-assisted or telerobotic surgery, the surgeon typically operates a control device to control the motion of surgical instruments at the surgical site from a location that may be remote from the patient (e.g., across the operating room, in a different room or a completely different building from the patient) or immediately adjacent to the patient. The controller usually includes one or more manually-operated input devices, such as joysticks, exoskeletal gloves or the like, which are coupled (directly or indirectly) to the surgical instruments with servo motors for articulating the instruments at the surgical site. The servo motors are typically part of an electromechanical device or surgical manipulator that supports and controls the surgical instruments that have been introduced directly into an open surgical site or through trocar sleeves into a body cavity, such as the patient's abdomen. During the operation, the surgical manipulator provides mechanical articulation and control of a variety of surgical instruments, such as tissue graspers, needle drivers, electrosurgical cautery probes, etc., that each perform various functions for the surgeon, e.g., holding or driving a needle, grasping a blood vessel, or dissecting, cauterizing or coagulating tissue. A surgeon may employ a large number of different surgical instruments/tools during a procedure.
Robotic manipulators, particularly small ones of the type used in minimally-invasive invasive surgery, generally need to transmit actuation forces and sensor information some distance through the structure of the robotic device, because the mass and size of presently-available actuators (e.g., motors) and sensors (e.g., position encoders) are too large to permit them to be directly located where the force and sensing is needed. In a mechanical device of a type used in minimally-invasive surgery, it is desirable to provide (a) significant force and (b) positioning precision at the tip of a small structure. However, the motors and sensors used are typically too large and heavy to be placed inside the patient's body. In addition, safety concerns make transmission of electrical power to actuators or sensors located inside the patient's body problematic. Therefore, in these robotic applications, it is desirable to transmit the actuation forces for multiple degrees of freedom over significant distances, and the desire to minimize the size of incisions places a premium on the cross-sectional area used for the structure and the transmission of the actuation forces. In addition, it is desirable to keep the transmission mechanism for these forces as stiff and friction-free as possible, so that the proximally sensed position of the proximal actuator may be used as a fair representation of the distally positioned joint.
There are many means known to transmit force over a distance. Some examples include: cables in tension; rods or tubes in compression; torsion; and hydraulic or pneumatic pressure. None of these force transmission means are ideal for minimally-invasive surgery, and therefore there are practical limits to the precision with which a distally mounted structure member can be moved, and the force it can provide. For example, cables may stretch, thereby leading to errors in the distal structure position relative to the position of the proximal driving mechanism. Compliant cables may also have significant friction with the supporting structure. Friction is notoriously difficult to model, may behave in a time-varying way and is dependent upon any number of variables, including the position (and past history of the position) of any intermediate joints in the structure, and when combined with compliant actuation can lead to stick-slip behavior. These errors limit the utility of the robot, limit the number of degrees of freedom that can be controlled remotely through a small structure, limit the use of the robot's state to provide haptic feedback, and are particularly severe for a flexible surgical robot which must accurately control many degrees of freedom of actuation at the end of a long, flexible structure.
Feedback regarding the actual position of a structure is a commonly applied mechanism for compensating for these errors. If the position of the distal part is accurately measured with minimal delays, control algorithms can compensate for many of the deficiencies of the force transmission mechanisms. While there are limits to the compensation that can be done with feedback, those limits cannot even be approached without accurate, timely information about the positions of the distally-mounted parts. The main restriction on the applicability of feedback in these types of robotic arms is the lack of an effective means of determining the position of parts mounted at the distal end of the structure.
Two methods of determining the position of a structure are proprioception and exteroception. Proprioception refers to the internal sensing of the position of the structure and exteroception refers to the external sensing of the position.
A proprioceptive system may monitor the motors actuating the movement of each joint in a robotic structure. By monitoring the degree of movement of each motor, the expected movement of each corresponding joint can be estimated. For example, the da Vinci Surgical System by Intuitive Surgical, Inc., of Sunnyvale, Calif. utilizes motor sensors (e.g., encoders) for position feedback. The encoders are co-located with the actuators, which are positioned outside of the patient body and drive the structure's joints through a cable mechanism. However, this method allows the position control loop to compensate only for error sources that occur between the actuator and the encoder. Error sources distal to the encoder are not observed by the encoder and thus cannot be compensated. This arrangement does not provide for compensation of errors introduced by any structures or drivetrain mechanisms that are distal to the encoders. In order to estimate the position of distal joints, the cable mechanism is modeled as infinitely stiff, so the motor position is assumed to be a reliable indication of the actual joint position. If the drivetrain between the actuator and the joint is not sufficiently stiff or other errors exist between the joint and the encoder, there will be a difference between actual joint orientation and the expected orientation based on the motor position. These joints are often mounted serially, so that the relative orientation of each of the links in the structure must be known and transformed to determine the position of the distal end of the structure. As a result, errors in the sensing of the orientations of intermediate links may compound along the way.
In other cases, position sensors may be positioned directly at the joints. This arrangement may be effective for larger structures, such as the Canadarm2 used on the International Space Station, and industrial robots. In robotically-assisted surgical systems having many degrees of freedom near the distal end, joint encoders do not offer much benefit as they are too large to be placed where the most problematic joints are located, and the connections required to transmit their position data back to the controller compete for space with the force-transmitting apparatus inside the arm.
Tachometers and accelerometers have been used, typically to estimate the robot's configuration at successive time steps and for force-control applications. These systems also do not provide complete position data and therefore cannot be used to compensate for some types of errors described above.
In systems which utilize exteroception, externally observed position feedback may be used to determine the position of a robotic structure.
For example, GPS-based location (especially differential GPS) has been used in outdoor environments, but typically can only be applied to large structures due to the relatively poor spatial resolution.
Field-based sensing can be accomplished using either AC or DC magnetic fields. In this approach, a small sensing element is placed at the tip of the structure to be monitored, and a magnetic field generated externally. While these systems can achieve good accuracy under ideal conditions, the accuracy rapidly degrades when metallic objects are nearby (as is the case in most surgical applications), and an external apparatus is needed to generate the field. In addition, these sensors do not directly encode the positions of the joints or links in a structure unless a sensor is provided on each link. Therefore, different configurations of the structure which result in the same position of the sensor are not distinguishable, which limits the use of this type of data for feedback compensation as described above.
Another approach is to utilize a dedicated set of unloaded cables, attached to the movable member at the distal end and to an encoder or position measurement device at the proximal end. Although unloaded, these cables are still subject to the same bending at the intermediate joints as are the actuation devices, and to provide feedback for many degrees of freedom, many cables must be used, requiring a larger structure.
Methods have been used involving a combination of sensing of the external environment, locating landmarks, and using this information to simultaneously construct a map of the environment and localize the robot. However, these systems are only applicable to larger and slower-moving systems, such as mobile robotic platforms. In the surgical applications considered here, these methods may be undesirable because the landmarks (e.g. patient anatomy) are not well defined and may change position as a result of disease states or during the surgical manipulations.
Techniques known as visual-servoing have been proposed, in which a camera mounted on the end of the robot arm is used in combination with joint encoders to control the position of the robot end-effector with respect to a tracked object in the field of view of an imaging device. This approach suffers from the need to provide joint position feedback and identifiable landmarks in the viewed scene, which are also problematic in surgical applications.
These techniques suffer from various deficiencies when utilized in robotic surgical applications and do not provide the type of information desired for implementing closed-loop control of such robotic mechanisms and compensating for the deficiencies in the force-transmission means described above. In addition, these techniques may be complex and expensive to implement.
Accordingly, it would be desirable to provide systems and methods for determining the position of a surgical instrument at a surgical site on a patient. In particular, it would be desirable for these systems and methods to provide real-time position feedback to the control system for a robotic surgical instrument.